[Thuban-commits] r2688 - in trunk/thuban: . Doc/manual Resources/XML Thuban/Model Thuban/UI test

[Frank Koormann]

Am Freitag 30 Juni 2006 15:23 schrieb Bernhard Herzog:
> > New Classification "Pattern": Classify text attributes by regexp.
> Nice.
Thanks.

> However, do we really want full regular expression matches?  I 
> ask because we once thought about moving to Styled Layer Descriptor
> (SLD, [1]) syntax for the classifications.  AFAICT regular expressions
> are not supported there.  What SLD has, though is a simple wildcard
> matcher (PropertyIsLike, [2]).
Hm, SLD reads to be made for more or less static visualisation. Since Thuban 
is a tool for interactive exploration of data I would miss on-the-fly 
grouping like "[A-M].*". Following the SLD way I would have to preprocess the 
data to add a (temp?) column

> > Modified: trunk/thuban/Resources/XML/thuban-1.1.dtd
> > -<!ELEMENT classification (clnull?, clpoint*, clrange*, clcont*)>
> > +<!ELEMENT classification (clnull?, clpoint*, clrange*, clpattern*,
> > clcont*)>
>
> There seems to be a bug in the DTD here.  It should probably read
> <!ELEMENT classification (clnull?, (clpoint | clrange | clpattern |
> clcont)*)>
I simply extended the existing line. Someone with more knowledge on DTDs, 
please comment and file a bug if necessary.

> BTW, does anyone know what clcont is supposed to mean?  It only occurs
> in the DTD but nowhere in the code or tests and it has apparently been
> in the DTD ever since we had a DTD afaict.
No idea where this comes from. I think we should remove it.

> > +class ClassGroupPattern(ClassGroup):
> I would prefer a __repr__ that does not look like that of a tuple.  It
> should make clear that it is the repr of an instance of
> ClassGroupPattern.  This is a bug in all the classes in that module,
> though.
I added a bug and picked it up.

> > Modified: trunk/thuban/test/test_classification.py
> Shouldn't at least one test actually use a regular expression that's not
> the same as a string equality test?

It does test two times for "first character", not string equality: 
| TestClassGroupPattern.test
| [...]
|       group.SetPattern("A")
|        self.assertEqual(group.GetPattern(), "A")
|
|        # test Matches()
|        self.assertEqual(group.Matches("CBA"), False)
|        self.assertEqual(group.Matches("ABC"), True)

and
|TestClassification.test_multiple_groups_textual        
|         p = ClassGroupPattern("A")                
|         self.assertEquals(self.clazz.FindGroup("Abc"), p)

Regards,

	Frank

-- 
Frank Koormann                             <frank.koormann at intevation.de>
 Professional Service around Free Software       (http://intevation.net/)
 FreeGIS Project                                 (http://freegis.org/)